Document clustering using locality preserving indexing

By Deng Cai, Xiaofei He, Jiawei Han. IEEE Transactions on Knowledge and Data Engineering, 2005

800 citations recorded by the source corpus.

We propose a novel document clustering method which aims to cluster the documents into different semantic classes. The document space is generally of high dimensionality and clustering in such a high dimensional space is often infeasible due to the curse of dimensionality. By using locality preserving indexing (LPI), the documents can be projected into a lower-dimensional semantic space in which the documents related to the same semantics are close to each other. Different from previous document clustering methods based on latent semantic indexing (LSI) or nonnegative matrix factorization (NMF), our method tries to discover both the geometric and discriminating structures of the document space. Theoretical analysis of our method shows that LPI is an unsupervised approximation of the supervised linear discriminant analysis (LDA) method, which gives the intuitive motivation of our method. Extensive experimental evaluations are performed on the Reuters-21578 and TDT2 data sets.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/b2746adae0413e87aec01303c2d8a2a67113bf11

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

Comments on "Globally Maximizing, Locally Minimizing: Unsupervised Discriminant Projection with Application to Face and Palm Biometrics" (2007, background): “…the mean square distance from any points to the global centroid, i.e., Similarly, LPP characterizes its global quantity as the weighted sum of square distances from any points to the origin, i.e., According to [5], the weighted mean of sample set, denoted as should be removed before LPP is applied.”
Graph Regularized Sparse Coding for Image Representation (2011, methodology): “Two standard clustering metrics, the accuracy (AC) and the normalized mutual information metric are used to measure the clustering performance [51].”
Deep Comprehensive Correlation Mining for Image Clustering (2019, methodology): “The clustering accuracy is deﬁned by: where and function map( x ) denotes the best permutation mapping function gained by Hungarian algorithm [5].”
Non-negative Matrix Factorization on Manifold (2008, methodology): “Two metrics, the accuracy (AC) and the normalized mutual information metric (MI) are used to measure the clustering performance [29][5].”
On Unifying Multi-view Self-Representations for Clustering by Tensor Multi-rank Minimization (2016, methodology): “thod is shown in Fig. 5, where row and column names are true and predicted labels respectively. Here, the cluster label is predicted by the best permutation mapping function used in the metric of ACC [33]. We can see that, compared with LTMSC, the proposed method wins in almost all categories in ter”
Locally Consistent Concept Factorization for Document Clustering (2011, methodology): “Thus, we call it Locally Consistent Concept Factorization (LCCF).”
A Similarity Measure for Text Classification and Clustering (2014, background): “The accuracy, AC , and entropy, En , are adopted to gauge the clustering performance [10], [52] as follows: where Most i is the majority number of documents having identical class labels in the i th cluster, n i is the number of documents in the i th cluster, and n is the number of documents with…”
Coupled nominal similarity in unsupervised learning (2011, methodology): “The clustering performance is evaluated by comparing the obtained cluster of each object with that provided by the data label in terms of accuracy (AC ) and normalized mutual information (NMI ) [3].”
An Entropy Weighting k-Means Algorithm for Subspace Clustering of High-Dimensional Sparse Data (2007, background): “Clearly, clustering of high-dimensional sparse data requires special treatment [2], [3], [4], [5].”
Large Scale Spectral Clustering Via Landmark-Based Sparse Representation (2015, neutral): “com/exdb/mnist/ information (NMI) metric [3] to measure the clustering performance.”
Large Scale Spectral Clustering with Landmark-Based Representation (2011, methodology): “We use the accuracy (AC) (Cai et al. 2005) to measure the clustering performance.”
Learning a Kernel for Multi-Task Clustering (2011, methodology): “So we define the adjacency matrix W as follows (Cai, He, and Han 2005), Wij = { cosine(xi,xj), if xi,xj ∈ Xk and xj ∈ N (xi) or xi ∈ N (xj) 0, otherwise.”
Graph Regularized Nonnegative Matrix Factorization for Data Representation (2011, background): “Please see [4] for the detailed definitions of these two metrics.”
Self-Taught Convolutional Neural Networks for Short Text Clustering (2017, methodology): “Following previous study [39], the number of nearest neighbors in Eqn.”